Services Watchdog

ReviewAudited by ClawScan on May 16, 2026.

Overview

The watchdog idea is legitimate, but this package ships a personalized script that can run continuously and use a local Telegram bot token to notify a hard-coded chat.

Do not enable this skill unmodified. First replace the hard-coded services and paths, remove or explicitly configure Telegram notifications, inspect or create the missing systemd service/timer files, and make sure you understand that loginctl linger keeps the watchdog active after logout.

Findings (5)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

Medium

#ASI09: Human-Agent Trust Exploitation

What this means

If installed without careful editing, it may monitor and start the publisher's named service layout instead of only services the user explicitly chose.

Why it was flagged

The package is described as a reusable watchdog for Node.js services, but the actual shipped script defaults to specific personal projects rather than a generic template.

Skill content

# שלושת השירותים:
#   1) sahi-diet      → ~/projects/sahi-diet/src/bot.js
#   2) sahi-mind      → ~/projects/sahi-mind/src/index.js
#   3) mission-control→ ~/projects/mission-control/server.js
...
services=(diet mind mc)

Recommendation

Publish this as a clearly marked example or replace the defaults with user-configured service names and paths before enabling the timer.

Medium

#ASI03: Identity and Privilege Abuse

What this means

A local Telegram bot credential could be used to send messages to a chat the user did not configure.

Why it was flagged

The script reads a Telegram bot token from a local .env file and pairs it with a hard-coded chat ID, rather than requiring an explicit user-provided notification target.

Skill content

token=$(grep -E '^TELEGRAM_BOT_TOKEN=' "$WORKSPACE/projects/sahi-diet/.env" ...)
local chat_id="6034574482"  # David

Recommendation

Remove this default notification code or require the user to explicitly configure the bot token and chat ID; declare this credential use in metadata.

Medium

#ASI07: Insecure Inter-Agent Communication

What this means

Service names and recovery/failure events may leave the machine and be delivered to a hard-coded Telegram recipient.

Why it was flagged

Watchdog status messages are sent to Telegram using a fixed destination; the main SKILL.md does not clearly disclose this external data flow.

Skill content

curl -s --max-time 10 -X POST "https://api.telegram.org/bot${token}/sendMessage" \
  -d "chat_id=${chat_id}" \
  -d "text=${msg}"

Recommendation

Make outbound notifications opt-in, document exactly what is sent, and require the user to set the destination.

Low

#ASI10: Rogue Agents

What this means

The watchdog can keep running and restarting services after SSH disconnects, logout, and potentially reboot.

Why it was flagged

The skill intentionally installs a persistent user-level timer that continues after logout; this is aligned with a watchdog, but users need to understand the persistence.

Skill content

systemctl --user enable --now sahi-watchdog.timer
loginctl enable-linger "$USER"   # keeps the timer running when not logged in

Recommendation

Only enable it after reviewing the script; document disable steps such as systemctl --user disable --now for the timer and loginctl disable-linger if appropriate.

Low

#ASI04: Agentic Supply Chain Vulnerabilities

What this means

The persistent systemd setup cannot be fully reviewed from the included artifacts and may not install as documented.

Why it was flagged

SKILL.md references scripts/sahi-watchdog.service and scripts/sahi-watchdog.timer, but those unit files are not present in the reviewed package.

Skill content

2 file(s): SKILL.md (5135 bytes); scripts/services-watchdog.sh (4595 bytes)

Recommendation

Include the unit and timer files in the package, or require users to create and inspect them explicitly before enabling the watchdog.